On the convergence of a weak greedy algorithm for the multivariate Haar basis (Q485308)

The paper is concerned with greedy approximation. For history and background on greedy approximation the recent book by \textit{V. Temlyakov} [Greedy approximation. Cambridge: Cambridge University Press (2011; Zbl 1279.41001)] is highly recommended. \textit{S. J. Dilworth} et al. [J. Approx. Theory 114, No. 2, 214--241 (2002; Zbl 1018.46006)] proved that the so called Tresholding Greedy Algorithm (TGA) with respect to the normalized Haar basis in \(L^1[0,1]\) could not converge for some initial vectors. \textit{S. Gogyan} [J. Approx. Theory 161, No. 1, 49--64 (2009; Zbl 1177.41032)] proposed a version of this algorithm, called weak tresholding greedy algorithm (WTGA) for the Haar basis in \(L^1[0,1]\) which converges. The aim of this paper is to extend Gogyan's result to multivariate Haar system in \(L^1[0,1]^d\). As the authors mention ''The first obstacle in extending the one-dimensional algorithm, which impeded progress on this problem for some time, is that the ``obvious generalization'' fails to converge. Therefore, a more complicated algorithm is required, which depends on two parameters.'' A consequence of the obtained results is that the TGA algorithm for the multivariate Haar basis in \(L^1[0,1]^d\) cannot converge for some initial functions.